Turbopump Fault Detection Algorithm Based on Wavelet RMS Vectors and SVM

Abstract

A fault detection algorithm based on wavelet RMS (Root Mean Square) vectors and SVM (Support Vector Machine) is proposed for LRE(Liquid Rocket Engine) turbo pump real-time fault detection. The algorithm divides the historical signals into some segments by reasonable step length and gets M-layer approximation signals and M-layer detail signals through Daubechies wavelet decomposition and reconstruction of the segments. Then it divides every layer into K average segments and calculates RMS of each segment, gets M RMS sequences of approximation signals and M RMS sequences of detail signals, constructs 2M-dimensional RMS vector as fault feature by extracting RMS values at the same position in every RMS sequence, and extracts all the fault feature vectors of historical signal to construct SVM training sample set and then obtains SVM classifier for real-time fault detection of the test signal. A part of the vibration acceleration signal of a certain type of turbo pump was chosen as the test object to validate the algorithm. The test results showed that for the test signal within 20.80s' duration, the algorithm detected the faults and gave an alarm at 20.20s without false alarm and missing alarm, and there was a delay of 0.19s (less than 0.5s) after the faults really occurred. The algorithm has good accuracy and real-time performance.